The Kolmogorov-Smirnov’s Test for Authorship Attribution on the Phonological Level
Iryna Khomytska, Vasyl Teslyuk, L. Bordyuk · 2020 IEEE 15th International Conference on Computer Sciences and Information Technologies (CSIT) · 2020
In the conducted research the experiments have been made with parametric and non-parametric texts for text differentiation. The Kolmogorov-Smirnov’s test has proved the most efficient in authorship identification. The test is more powerful than the Student’s t-test and the chi-square test. The non-parametric two-sample test for revealing significant differences between texts by different authors has allowed us to differentiate the texts of the publicist and the fiction styles. This test has been done in eight consonant groups showing significant differences in all of them. Consequently, the phonological language level is optimal to differentiate styles on. The success rate equals 97%, 98%. The software system has been developed to distinguish between texts by different authors. The cross platform Java programming language has been used. The transcribed texts are stored in the program and make it independent of the Internet.